Proceedings of the 11th International Ship Stability Workshop
An Approach to the Validation of Ship Flooding Simulation Models
Egbert L. Ypma
MARIN, the Netherlands
Terry Turner
Defence Science & Technology Organisation (DSTO), Australia
ABSTRACT
A methodology has been developed to validate a Ship Flooding simulation tool. The approach is to
initially validate the flooding model and the vessel model separately and then couple the two
models together for the final step in the validation process. A series of model tests have been
undertaken and data obtained has been utilised as part of the validation process. Uncertainty in the
model test measurements and the geometry of the physical model play a crucial role in the
validation process.
This paper provides an overview of the methodology adopted for the validation of the ship flooding
simulation tool and presents some of the preliminary results from this study.
KEYWORDS
Time domain, flooding, simulation, damaged stability, validation, uncertainty determination
.
INTRODUCTION
To accurately predict the progressive flooding
of a damaged vessel and its effect on the ships
motion two tightly coupled methods are
required to be developed. If the vessel changes
its orientation, the internal floodwater
distribution changes and vice versa. In
addition, the changing distribution of the
floodwater changes the dynamics of the vessel
(centre of gravity, total mass and mass inertia).
All these effects have to be taken into account
in the modelling process.
An added complexity in trying to accurately
simulate the flooding phenomenon is the highly
non-linear chaotic nature of the flooding
process. Small variations in this flooding
process, e.g. how the water progresses through
an opening, can influence the final result.
Due to the highly chaotic nature of the flooding
process it is vital that the numerical model
represents the experimental model as closely as
possible. To obtain an exact numerical
representation of the physical model is very
difficult. Differences may occur due to the
limited accuracy of the production process of
the physical model, modelling errors made in
the translation from „real‟ world to „simulated‟
world, and the uncertainty (or limited accuracy)
of the measurements. All these factors must be
considered.
When there is a requirement to numerically
model an actual full scale vessel other factors
Proceedings of the 11th International Ship Stability Workshop
also must also be considered. The internal
geometry of a full scale ship is extremely
complicated and it would be very difficult to
account for all the small details that may
influence the flooding process. Issues such as
leaking doors and collapsing air ducts are
highly random events that can never be
accounted for numerically. For this reason
several “acceptable” assumptions are required
to be made when numerically modelling full
scale ships.
Due to the highly chaotic nature of the flooding
process and the various areas of uncertainties in
both model scale and full scale vessels, a
method for progressive flooding tools must be
developed to account for these uncertainties at
an acceptable level. The following three
questions need to be carefully considered when
defining this method:
1. How can a validation process be defined
such that it is possible to conclude whether
a simulation tool is sufficiently accurate?
2. Is it possible to define general rules to
model the internal ship-geometry in such a
way that the simulation tool predicts
extreme events sufficiently accurate (both
statistically and in magnitude)?
3. What is the best way to deal with
uncertainties in the validation process?
This paper will provide a brief overview of the
numerical tool Fredyn [see Fredyn v10.1 2009]
and its progressive flooding modelling
capability and will also outline the approach
undertaken by both The Maritime Institute
Netherlands, (MARIN), and The Defence
Science and Technology Organisation, (DSTO)
for the validation of the progressive flooding
module.
SHIP MOTION AND PROGRESSIVE
FLOODING SIMULATION MODEL
Background
The Cooperative Research Navies group,
(CRNav), was established in 1989 to initiate a
research program focussed on increasing the
understanding of the dynamic stability of both
intact and damaged naval vessels. The group
has representatives from Australia, Canada,
France, The Netherlands, the United Kingdom
and the United States. The CRNav aims to
increase the understanding of the stability of
Naval vessels from a more physics based
approach rather than an empirical derived one.
To manage this process the CRNav has formed
the Naval Stability Standards Working Group
(NSSWG). The objective of the NSSWG is to
investigate the applicability of the quasi-static,
empirical based Sarchin and Goldberg stability
criteria for modern Navy vessels and to
develop a shared view on the future of naval
stability assessment.
Until recently, the main focus of the NSSWG
has been on intact stability and so far this has
been a fairly comprehensive and complex task.
However with the recent development of the
new flooding module within Fredyn, future
programs of work will be focused on the
damage stability of naval vessels.
The flooding simulation model was developed
and implemented by MARIN and funded by
the CRNav group. In 2009 a collaboration
agreement was signed between the CRNav and
the Cooperative Research Ships, (CRS),
working group ShipSurv II to jointly develop
and validate the flooding module. In 2009 and
2010, the Defence Science and Technology
Organisation,
(DSTO),
Australia,
in
collaboration with The Australian Maritime
College, (AMC), have undertaken a research
program to support MARIN in the validation of
the progressive flooding module.
In theory the flooding module can be interfaced
to any 6D vessel (large) motion simulation
Proceedings of the 11th International Ship Stability Workshop
program. Currently, however, it is interfaced to
Fredyn, jointly developed by MARIN with the
CRNav, and PRETTI (jointly developed by
MARIN and the CRS group). Fredyn was used
for all the examples in this paper.
The Simulation Model
To enable an accurate simulation of the
flooding of a damaged vessel operating in
waves the simulation model describing the
motions of the vessel and the model that
determines
the
progressive
flooding
mechanism must be closely coupled to each
other. Figure 1 shows an example of the typical
information that is required to be interchanged
between the simulation models..
Total Mass,
CoG & Inertia
Vessel Simulation
Model
Flooding Module
Vessel
Motions
Figure 1 Interface Vessel model and flooding model
In the scenario with significant flooding
onboard a vessel, it is not unusual for sudden
large changes in the vessels motions to occur.
For an accurate simulation of this event both
the flooding model and the vessel motion
model must take this into account. The
simulation must also be able to calculate a
changing mass, centre of gravity and inertia
over time. The accuracy of the roll damping
model utilised is also vital in modelling this
scenario. The relative wave height at the
damage location is also required to be
accurately predicted.
FREDYN©
FREDYN© is an integrated sea keeping and
manoeuvring ship simulation tool capable of
predicting large ship motions in extreme
conditions. The development of FREDYN was
jointly funded over the last 20 year period by
MARIN and the CRNav. Over the years a
substantial effort was made to validate and
improve the code with model test experiments.
FREDYN uses the frequency domain tool
Shipmo2000 as a pre-processor to calculate the
frequency dependent added mass and damping
coefficients. Using a panellised hull form the
program is capable of calculating the FroudeKrylov forces on the instantaneous wetted hull.
An appropriate roll damping model can be
selected which can be tuned to satisfaction
when roll decay data is available. The standard
vessel manoeuvring model is based on a frigate
hull form but can be replaced with a dedicated,
specifically tuned model when required. A
wide variety of simulation components is
available to model the vessel‟s propulsion and
manoeuvring: rudders, fins, skegs, bilge keels,
various propeller types, waterjets, trim-flaps,
etc. The forces calculated by these sub-models
are partly empirical based. A separate module
is used to control the vessel‟s heading and/or
attitude.
The environment can be modelled by wind and
by various wave systems coming from different
directions. The wind and wave systems can be
specified by making a selection from one of the
available spectra. Together with the user
specified parameters this will generate a
(random) wave sequence (or varying wind
speed).
The equations of motions are based on
Newton‟s second law: the forces and moments
of all the sub-models are calculated, transferred
to the space-fixed reference frame attached to
the ship‟s centre of gravity, and summed. It
eventually results in the momentum rate.
Fredyn allows for time-varying mass properties
to be able to deal with the potentially large
mass fluctuations caused by the flooding
process.
To increase the flexibility a (python) scripting
module allows to interact with the simulation
program without altering its code. The
scripting functionality can be used to program
Proceedings of the 11th International Ship Stability Workshop
vessel trajectories, implement speed pilots or
do more advanced logging.
Over the last year the software was completely
restructured resulting in a very modular and
highly configurable simulation program that
can be very easily extended by additional
modules. More recently, efforts are made to
incorporate the steady forward wave and the
diffracted and radiated wave patterns to
improve the calculation of the wave profile
close to the moving hull.
Flooding Module
The flooding module is used to calculate the
flow of water and air through a user specified
geometry. It assumes a horizontal fluid surface
at all times. The effect of air-compressibility
and its effect on the flow of water is fully taken
into account.
The compartment geometry is represent by
tank-tables that are generated prior to the
simulation. A tank-table for a compartment
tabularises the relation between heel, trim of
the vessel, level of the fluid in the compartment
and the volume of the fluid, centre of gravity
and inertia matrix. Interpolation on actual heel,
trim and level values is used to find
intermediate values.
Any number of openings can be specified
connecting two tanks or a tank to the sea. A
single opening consists of four corner-points
that specify the size and orientation of the
opening. For each opening a constant discharge
coefficient for water and a separate discharge
coefficient for air has to be specified. If
required, the user can specify a leaking
pressure and area, a collapse pressure and a
start- and/or stop time. An opening either
connects two tanks or connects a single tank to
the sea.
It is also possible to define a duct between two
tanks, or between a tank and the sea.
Bernoulli‟s equation for incompressible media
is used to determine the flow velocity of fluid
along a stream line from the centre of a
compartment (A) to the opening (B)
pB
1
.
2
pA
w
.vB2
g.
w
. hB hA
0
(1)
In this application of Bernoulli‟s equation the
is constant and the velocity in point A, the
centre of the compartment, is neglected. The
variables p A and p B are the air pressures above
the fluid in the compartment (A) and on the
other side of the opening (B). After
determining the velocity in the opening the
mass flow through an entire opening is
determined by integration over the height of the
opening (along the local vertical):
H
mw
w
.Cd , w . width z .vB z .dz
(2)
0
Where C d is the discharge coefficient specified
by the user for this opening. It takes all the
losses into account caused by contraction,
pressure losses etc.
A similar procedure is used to determine the
mass flow of air through an opening and the
Bernoulli‟s equation for compressible flow is
used:
p0
0
. ln pB
ln p A
1 2
.vB
2
0
(3)
The flooding process is considered isothermal,
hence Boyle‟s law applies and thus the density
of air is assumed to vary linearly with the
pressure.
The pressure correction method developed for
air and water flows by Ruponen [see Ruponen
2007] is used to solve the coupled flow of fluid
and air through a complex user defined
geometry. The pressure correction method is
using the equation of (mass)continuity and the
linearised equations of Bernoulli to correct the
water levels and air-pressures in an iterative
method until the error in mass flow drops
Proceedings of the 11th International Ship Stability Workshop
below a user specified minimum. Upon
conversion both the equations of mass
continuity and momentum are satisfied.
FREDYN FLOODING MODEL VALIDATION
As previously discussed it is extremely difficult
to model and capture all the phenomena that
occur during the flooding of a vessel. For this
reason there will always be slight variations in
the experimental results when compared to the
numerically predicted results. There are several
factors that contribute to the differences
observed. These include:
The issue with the uncertainty covered by the
physical model and measurements can be
solved by firstly having a clear understanding
of the uncertainties involved and secondly by
undertaking a series of simulations to
determine the influence that these uncertainties
have on the overall result.
The uncertainty in both the flooding and vessel
motion models can be solved by separating the
validation of the flooding model and the vessel
simulation model. The validation process can
then be split into several phases:
1. Fully constrained model.
Measurement accuracy (motions, levels)
Determination of the hydrostatic &
dynamic properties of the model
Production accuracy of the test model
(both internal & external)
Choices made during the modelling of the
internal geometry (deck and bulkhead
positions, permeability)
Imperfections caused by mathematical
modelling (empirics!)
These uncertainties can be grouped in 3 main
categories:
Uncertainties caused by the physical model
& the measurements
Uncertainties caused by flooding model
Uncertainties caused by the vessel model.
The first category plays a role during the model
testing, the second two are tightly coupled and
play a role during the simulations. The result of
the validation process will be a comparison
between the measurements and the simulation
data. In general, when they are „acceptably‟
close then the conclusion is justified that the
simulation application performs well. The first
key problem in view of the nature of the
process and the uncertainties that play a role is
how to define „acceptable‟. The second key
problem is that if the result is not „acceptable‟
then which sub-model has to be changed to
improve the result.
Fully constraining the model in a predescribed heel, trim and draft allows for a
check of the flooding module without the
dynamics of the vessel. By using a heel and
trim value different from zero the performance
of the flooding module for inclined openings
can be validated. In addition, the geometry of
the numerical model can be verified.
2. Force the measured motions from the
model test upon the flooding module.
In this validation process the motions
measured experimentally are prescribed onto
the numerical model and the water levels and
volumes in each compartment are determined.
These levels and volumes are then compared
to those obtained experimentally. If the flow
rate into a compartment is different between
the measured and the predicted, then the
opening coefficient(s) of the compartment
openings can be tuned. However, for complex
geometries this might become a very difficult
task. This process is a verification that the
flooding model is working correctly. This
approach is shown in Figure 1.
Proceedings of the 11th International Ship Stability Workshop
1
Modeltest
Measurements
Measured
Motions
2
Flooding Module
Criteria
Measured
Levels
Simulated
Levels
4. Close the loop and combine both models
3
Comparison
Performance
Conclusion
Figure 1 Flowchart showing the approach for the Prescribed
Motion Phase
3. Use the result of step 2 (a time-varying,
‘wandering’ mass of all the flood-water)
and use it to excite the dynamic vessel
model.
When the outcome of step 2 is summed it can be
replaced by a single, time-varying mass (having
a centre of gravity, inertia and a weight) that
moves over and throughout the vessel. This
approach is allowed as long as the vessel is only
excited by this wandering and changing
floodwater mass and will not be applicable
when the vessel is also subject to other external
forces such as waves.
1
Modeltest
Measurements
2
Flooding Module
Total Mass, CoG & Inertia
Criteria
Measured
Motions
during the model test. A schematic showing this
process is shown in Figure 2. It is a test of the
quasi-static flooding model approach and of the
interface between the flooding module and the
vessel simulation program.
3
Vessel Simulation
Model
Simulated
Motions
4
Comparison
Performance
Conclusion
Figure 2 Flowchart showing the approach for the Moving
Mass Phase
The outcome of this step (vessel motions) can
be compared with the measured vessel motions
The final step in the validation process is to
undertake a simulation with the complete model
i.e. vessel model and flooding model, and
compare the predicted levels and motions with
the measured modeltest data. A schematic
outlining this process is shown in Figure 3.
1
Modeltest
Measurements
2
Flooding Module
Total Mass,
CoG & Inertia
Measured
Motions
&
Measured
Levels
Criteria
3
Vessel Simulation
Model
Simulated
Levels
Simulated
Motions
4
Comparison
Performance
Conclusion
Figure 3 Flowchart showing the approach for the Full
Simulation Phase
Component & interface verification
Prior to using the measurements in this
approach, the validation method was verified
by replacing the model test measurements with
the data obtained from the fully coupled
system. The process flow of this test is shown
in Figure 1.
Proceedings of the 11th International Ship Stability Workshop
compartment arrangement at a later stage. The
compartment block was located aft of
amidships.
1
Flooding Module
& Fredyn
Simulated Motions I
2
testHarness &
Flooding Module
4
Comparison
Figure 5 A photograph of the generic destroyer model
Total Mass, CoG, Inertia
3
Moving Mass &
Fredyn
Simulated Motions II
Figure 4 Component Verification
The motions as calculated by step 1 and 3
should compare very well as modeling errors
play no role. It is a test of the interfaces and
coordinate transforms involved.
MODELTESTS
The model tests undertaken at AMC were
performed in two phases. Prior to undertaking
these experiments, it was expected that the tests
were going to be complex and the lessons
learned from Phase 1 would be incorporated
into the second phase of testing. The test
program was set up in such a way that the
complexity of the scenario was gradually
increased. A priori simulations were performed
to determine the most interesting loading
conditions. Prior to the flooding tests roll
decays at zero speed were undertaken, in both
damaged and intact condition. The roll decay
data was used to tune the roll damping of the
simulation model. Fully constrained model
tests were performed with a initial heel and
trim of zero. In later tests a series of runs were
performed with a range of non zero initial heel
and trim combinations.
The experimental model used, shown in Figure
5, was a generic destroyer design (scale 1:40,
length approx. 3.0 meters) with a detailed
internal geometry. The model was built in such
a way that the damaged compartment block
could be replaced with a more complex
Two
compartment
arrangements
were
constructed and are referred to as simple and
complex compartments. The difference
between the simple and the complex model was
the addition of (longitudinal and athwart)
gangways. These arrangements comprised of 3
decks, 19 compartments of which 6 had level
measurements and 2 had air pressure
measurements.
In
addition
to
these
measurements the 6 motions of the vessel were
also recorded along with both internal and
external video.
The flooding was initiated by puncturing a
latex membrane that covered the large damage
opening as seen in Figure 6.
Figure 6 The A photograph showing the generic destroyer
damage opening
During the model tests many precautions were
taken to reduce the uncertainty of the results.
After each run a calculation tool was used to
check the equilibrium levels in each
(measured) tank with respect to each other and
to the still water plane. After each run a check
of the (level) calibration was done. A run was
repeated when spurious results were suspected.
Proceedings of the 11th International Ship Stability Workshop
After the tests the model was carefully remeasured to ascertain the „as build‟ situation.
VALIDATION RESULTS
The validation methodology approach, as
previously described, has been applied to the
data obtained with the first phase of the model
test done at AMC. The results described in this
paper are preliminary due to the final set of
data (of phase II) not being available at the
time of the analysis. The validation was
performed in following steps:
1.
2.
3.
4.
5.
6.
7.
were performed, each with a different initial
angle. Only slight tuning of the radius of
gyration, kxx , value was required to have a
good resemblance of the measured and the
simulated data for all initial roll angles. A
typical comparison between the numerically
predicted and the experimentally obtained roll
decay is shown in Figure 8.
Component & interface verification
The vessel roll damping
Vessel hydrostatics
Fully Constrained
Forced Motions
Moving Mass
Unconstrained Analysis
All plots and other data are given in full scale.
Component & interface verification
The roll angle was selected as performance
indicator for this step. Both data sets lie on top
of each other, indicating a successful check.
Figure 8 Roll decay for initial angle of 15 degrees
Vessel hydrostatics & tanktables
To check the hydrostatics of the simulation
model, each tank was filled individually at
50%, 100% and at an intermediate value
determined by an unconstrained simulation run.
This was done both in FREDYN and
PARAMARINE©,
[see
PARAMARINE©
2009] the latter program was used to generate
the tanktables used in the flooding simulation.
The equilibrium values for heel, pitch, draft
and tank centre of gravity were compared.
Slight differences were found. After
completion of Phase 2, the actual dimensions
of the experimental model were verified and
slight differences were observed to those
dimensions used in the preliminary analysis.
Verification of the hydrostatics will be repeated
using the re-measured dimensions.
Figure 7 Component verification - Roll angle comparison
Fully Constrained
Vessel roll damping model
During the model test the model was fully
constrained at zero heel and trim. This test
gives an indication of the performance of the
flooding module. The influence of the motions
of the vessel on the flooding process is not
The roll decay data was used to tune the roll
damping model. All flooding tests were done at
zero speed hence only the roll damping for that
speed required to be tuned. Several roll decays
Proceedings of the 11th International Ship Stability Workshop
considered at this stage due to the model being
fully constrained in all six degree of freedom..
Figure 9 shows an example of both the
simulation and experimental time traces of the
water level in a compartment that floods
through a number of openings and other
compartments. The air pressure in this
compartment remains ambient as the
compartment is open to the outside. The
discharge coefficient for all the openings are
set to a default value of 0.58.
Forced Motion
For the forced motion analysis the
experimental model motions (heave, pitch roll)
were used to drive the flooding module. To be
able to do this a small test Harness application
was created which loads a file with motion data
and the flooding component. Figure 10 shows a
comparison between the simulation and model
tests results along with the expected
uncertainty.
The simulation of the flooding of this
compartment shows excellent comparison to
the experimental data. The arrival of the water
at the position of the probe, the filling rate of
the compartment and the final equilibrium level
are all predicted extremely well.
A full uncertainty analysis was undertaken.
which incorporated the uncertainties for level
and calibration measurement and uncertainties
in the as-build situation. The 95% values (2*σ)
used in the calculation of uncertainties are 2.0
mm for the level sensor accuracy and 4.0 mm
for the geometry uncertainty (model scale
values). Preliminary results from the model
tests suggest that the 2.0 mm might be too low.
The 95% range appeared to be around +/-0.20
m (full scale) and is indicated by the
unconnected dots shown in Figure 9.
Figure 9 Comparisons between predicted and experimental
compartment water levels.
Figure 10 Comparisons between predicted and experimental
compartment water levels.
There is a difference between the model test
and the simulation slightly outside the limits of
uncertainty. The difference is approximately
constant as soon as the equilibrium is reached.
The compartment which Figure 10 is referring
to is connected to the sea and is fully
ventilated, therefore at equilibrium, the
distance between the water level in the
compartment and the still water plane should
be zero.
Proceedings of the 11th International Ship Stability Workshop
Figure 12 Comparisons between predicted and experimental
compartment water levels
Figure 11 Distance to still-water plane (simulation)
Figure 11 shows the distance between the water
level within the compartment and the still water
plane over time. It is evident that the
simulation is predicting the water level within
the compartment accurately.
When the experimental data is used (level,
heel, pitch, heave and draft) and the
equilibrium value for the distance to the still
water plane is calculated, then the difference is
0.23 [m], full scale. It should be noted that after
undertaking this preliminary analysis the
physical experimental model was re-measured
and slight differences in the geometry details
used in this analysis were found. It is planned
to undertake this analysis again with the new
geometry details.
This compartment is located adjacent to the
damage opening. When the equilibrium
measurement for this compartment is used,
then the calculated difference between the
compartment water level and the still water
plane is 0.03 [m], which is well within the
estimated uncertainty limits.
Moving Mass
Using the results from the forced motion runs a
time record of the total flooding mass and its
center of mass can be determined. These values
are then used to excite the vessel model. The
roll, pitch and heave motions of the simulation
vessel can then be compared against the
modeltest values. Figure 13 shows the
comparison between the experimental and
simulated results.
Figure 12 shows the water level within a
different compartment over time for both the
simulation and model tests.
Figure 13 A comparison between the roll versus time for
both the simulated and experimental analysis.
Proceedings of the 11th International Ship Stability Workshop
The initial numerically predicted roll angle is
larger than the measured value. This result is
consistent with that shown in Figure 8. This
compartment is on the port side of the vessel.
The simulation predicts a smaller volume,
(mass), of water in this compartment hence
resulting in an increase roll to starboard, (roll
angle is positive to starboard). Initially, also
some fluctuations are visible. This is an
indication that the experimental model has
more internal damping. It will also be caused
by the quasi-static approach in the flooding
module for this highly dynamic model test.
Figure 14 shows an good agreement between
the numerically predicted magnitude of pitch
compared to that obtained experimentally.
Figure 15 A comparison between the roll versus time for
both the simulated and experimental analysis
It is evident that there is a significant difference
between the numerically predicted and the
experimental results. As stated previously, the
comparisons shown in this paper are using the
assumed geometry arrangement but the remeasuring of the model upon completion of the
experimental phase has shown some slight
variations in the assumed geometry. These
slight variations at model scale may have
significant effect on the results when modelling
full scale. It is planned to undertake the
complete validation process again using the
updated geometry.
Figure 14 Comparison of pitch motion
Unconstrained run
The final stage in the validation methodology
is the complete coupling of the numerical
flooding model with the numerical vessel
model. Figure 15 and Figure 16 show a
comparison between the numerically predicted
roll and pitch motions compared with the
experimentally obtained values.
Figure 16 A comparison between the pitch versus time for
both the simulated and experimental analysis
Proceedings of the 11th International Ship Stability Workshop
CONCLUSIONS
This paper has provided an overview of the
methodology that has been adopted by MARIN
and DSTO to validate the progressive flooding
modelling capability within Fredyn. This
methodology included a phased approach
where initially both the flooding model and the
vessel model were validated separately and
then coupled together for the last validation
stage. A series of model tests were undertaken
in support of this validation process.
Preliminary findings have shown reasonably
good results but have also highlighted the need
to clearly identify the source and extents of
uncertainties in the experimental program. This
information can then be utilised to determine
what is an “acceptable” level of agreement
between the numerical predictions and the
experimental results.
The results presented in this paper are based on
an assumed geometry of the experimental
model and post trial geometry verification have
shown slight differences in the geometry
details. Ongoing validation studies are planned
using the new geometry definitions.
REFERENCES
Fredyn v10.1, Theory & user Manual, MARIN, the
Netherlands, 2009
PARAMARINE© v6.1, User Manual, GRC Ltd, United
Kingdom, 2009
Ruponen, Pekka, Progressive Flooding of a Damaged
Passenger Ship, Doctoral Dissertation, Helsinki University
of Technology, 2007